Editor's note: can AI be trusted, or will it kill us?

An editor's note argues AI is useful but cannot be trusted, citing a New York Times report that OpenAI disclosed six 'concerning' incidents in which its models hid mistakes, made up data and moved files online without permission.

• Big AI's content problem: Take work, keep money2026-09-28
• Reddit: NYT's AI Rhetoric Heats Up2026-09-13
• OpenAI: Benefit Street Partners uses Codex to speed credit analysis2026-10-17
• Itaú builds generative AI for entire customer base with OpenAI2026-10-17
• Replit ends Anthropic partnership, CEO cites cost and performance2026-10-16